How does Tardigrade avoid clusters of nodes being correlated to physical location of SNOs?

There is no way to differentiate between a single SNO having multiple IPs and just a bunch of SNOs running nodes on multiple IPs. What exactly would you give a disadvantage to?

It’s not so much about being fair. I would say it’s totally fair to give SNOs that provide a faster service some advantages. But the network doesn’t really need that and giving any advantage for something that isn’t really needed would lead to a less decentralized spread of data.

Besides, a bad ISP can give you a lot more latency than a good ISP + a good VPN would. Non of it is a magic bullet solution. But perhaps combining all possible signals and training a model on too of all that data could lead to some good mechanisms. As long as the inference can still be lightning fast as node selection happens right during the transfer and the customer is actually waiting on it. Although perhaps some asynchronous scoring and pre-weighting the node selection process could be done.

1 Like

yeah good point… i guess in the end is just about data lol

it’s rare that solving one problem doesn’t create others… or reveal them…

i often find solution in the inverse of things… if nothing else radical different approaches often gives good clues to the problems or shows what absolutely doesn’t work…

relying completely on ip subnet’s means the more ip’s one has the more data one gets in a shorter time frame… anything that reduces that and any other advantage people get out trying to game the system, is a long term boon for the network…

it could be that if multiple illicit nodes are confirmed in one location, then those nodes are simply DQ or the entire location is DQ

it’s always a risk vs reward deal… right now my server could make double of what it is doing… and it would only costs me like 1/5th of what its making do to that… sure i add more overhead, but my bottom line would be much improved… but really thats just because the disks aren’t all full…

if i was lacking capacity then there wouldn’t be any point…
yes there is and will most likely never be a silver bullet for solving the issue… it will be a series of steps that make it not worthwhile… ofc there will always be the savy pirate here and there… but what problems are ever really solved…

we can build great big skyscrapers that reach the heavens… but really they are as impermanent as a sandcastle, to the flow of time… works for a while, but really the only thing that seems to be constant is change… and a few… “laws” which might be laws… but really when do we know we will never get smarter, and consider the wisdom of the past the babbling of idiots…

xD read ye who shall find in days unborn

1 Like

On what figure do you base this statement?

2 Likes

BTW, after thinking about this for a while: given enough nodes, nodes with VPNs should be outcompeted (on latency) by nodes without VPNs, so maybe this is not really a long-term problem?

2 Likes

I’m not sure about that. A typical VPN might have a higher latency but you could just rent a VPS or use some of your existing VPS and reroute the traffic. Latency will still be better than nodes that are just further away from the client.
Currently I’m using an ssh tunnel for my 2nd node and even though ingress is very low, both the local and the proxied node get the same amount of ingress. of course the proxy node is in the same country.

Also, what might be also interesting to know is how Storj plans to deal with people who consolidate their nodes that used to be in separate physical locations. There is at least one person stating here on the forum that they had several nodes in distinct physical locations, and they were moved to a single one, without using any VPNs or other SSH tunnels.

This case should be clear for satellites, as chunks that were available before in distinct /24 IP blocks are “suddenly” in the same /24 block.